GNOME Bugzilla – Bug 126443
CPU-bound looping seen in gnome-terminal
Last modified: 2006-02-14 06:38:50 UTC
Every so often, gnome-terminal starts consuming 100% CPU and stops responding to expose events. It doesn't seem to be triggered by anything in particular. I tried attaching gdb to a g-t process exhibiting this behaviour and getting some backtraces at various intervals. My guess is that it's a problem with vte rather than g-t.
+ Trace 41539
This one I only noticed once or twice. Of course, this is a rather hit-and-miss approach to finding the problem, as the sample set is quite small.
This is still happening with gnome-terminal 2.5.90. It seems to be particularly susceptible to the bug when I use the "screen" program over ssh, but that may be a coincidence. I also find when using screen that the display often becomes garbled and I need to request a redraw, which cleans it up. Sometimes, waiting or pressing Ctrl-PageUp repeatedly seems to fix it, but once it happens, it's usually not very long before it happens again, and before long g-t becomes unusable.
I'm seeing related stuff, which I've reported in bug #151680 ("gnome terminal crash"). I'm also seeing elevated CPU usage (often around 40% on P4 2600 cpu). I'm also using screen over ssh.
I ran a few tests. The CPU usage doesn't appear gnome-terminal dependent. It's just a matter of how much data is being dumped to the terminal, and seems the same as xterm. screen doesn't seem to make much difference. ssh does make some difference. A local "ls -R /" seems to generate half the CPU usage by the terminal than a remote (ssh) "ls -R /" does.
I have the same behaviour. The trigger for me is known, although not absolutely reproducible. It is the combination of gnome-terminal, screen and irssi. These three have a reasonable well known problem involving irssi rendering badly: basically it turns the entire terminal blue rather than rendering an IRC channel window. In my particular case, it also inspires this "consume 100% CPU for a long period of time" (long is between 20 seconds and 10 minutes, CPUs have been 667MHz PIII, 1.6GHz Celeron and 1.6GHz Centrino). I can generally cause it to happen within an hour of starting using irssi inside a screen session on gnome-terminal, but I haven't found a set of keystrokes that replicate it.
Can you still reproduce this using the latest release of vte?
I haven't observed this behaviour for some time, so feel free to close the bug for now.
I've not seen it for a long time either. And I've fixed a couple things recently that may have caused similar behavior. Closing for now. Reopen if you can observe with HEAD.